Group 1 (normals) and Group 2 (self-described empaths) are put in an fMRI and exposed to one of three kinds of stimuli. They're asked to do some sort of irrelevant, non-distracting task. Compare the three (four with control) scans, averaged, between groups.
[insert something clever]
1. Video of people watching a video that causes very strong emotions - with only their face in screen.
2. Actual person watchinig the same video (i.e. in the room, within sight of subject). Same restriction. (If sound is immportant for either, have them both wear headphones.)
3. Ditto, but with person near but not within sight of subject.
4. Nil (other than distractor).
5. (?) Subject watches same video.
It's obvious what I'm trying to get at, but this seems to be a fairly confounded design as stated. E.g. 1-2 vs 3-4, you drop visual stimulus - perhaps replace with some sort of neutral video? (I.e. visual stimulus would be the same between 3-4, but the person nearby would [maybe?] make a difference.)
If possible, would be good to have the person watching the video be someone close to the subject - or maybe have that only for the empaths, to give them an 'edge'.
Another confound is inter-task interference. Since every subject would have to be exposed to all four situations, and they'd have to be all the same emotional content (unless you wanted to do a two-variable study, and include e.g. one squick-provoking [yay medical/war videos], one warmth-provoking [puppies!], one .... (?), and one neutral [random documentary? art?]). And so you could have overlap, esp. since you'd probably have to do them at the same sitting. Randomization might help with this - perhaps have multiple trials e.g. 123423413412 if you have time.
Lastly, potential additional control needed - as mentioned in the previous list - of having them watch something neutral also. 4 would be 1 with neutral, but 2&3 would both need neutral video watching also.